What 50 jobs cost a US employer, and where AI already shows up in the work
Median wages from BLS for office and knowledge jobs, with benefits and payroll taxes added, pay in every state, and each job's core tasks checked against Anthropic's record of what people use Claude for.
Every figure is copied from a public source file, named on each page.
- Total compensation
- $37.64/h
- Wages
- $27.88/h
- Paid leave
- $2.34/h
- Ratio applied
- 1.2455
BLS Employer Costs for Employee Compensation, June 2026 (2026 Q2).
Benefits and payroll taxes add about 25% to what a US small business pays in wages, so a receptionist on the median $38,010 costs about $47,340 a year. Across these 50 jobs, 217 of 471 core tasks showed up in Claude.ai conversations in May 2026.
Office and administrative support
Sales
Business and finance
Management
Writing, design and media
Computer and IT
Legal
Health records
Where the numbers come from
The only arithmetic on these pages is the loaded cost: the wage times 1.2455.
- Wages
- BLS Occupational Employment and Wage Statistics, May 2025. Annual figures are the hourly wage times 2,080 hours, as BLS computes them, so they already include pay for paid leave.
- Fully loaded
- BLS Employer Costs for Employee Compensation, June 2026 (2026 Q2), private businesses with 1 to 49 workers: $37.64 an hour in total compensation over $27.88 in wages plus $2.34 in paid leave gives 1.2455. Paid leave sits in the denominator because the annual wage already pays it.
- Tasks
- O*NET 31.0 core tasks, the ten most important by their importance rating. A few occupations list fewer core tasks.
- AI use
- Anthropic Economic Index, release of June 2026, conversations from May 2026. "Did the task" is Anthropic's automation share, where the user directs Claude to do the work; "helped" is augmentation, where the user works with Claude to learn, check or iterate.
- Not measured
- What a particular business pays, and whether an AI system can do a task well. The AI figures count conversations about a task, not jobs replaced.
This page includes information from the O*NET 31.0 Database by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA). Used under the CC BY 4.0 license. O*NET® is a trademark of USDOL/ETA. AI use data from the Anthropic Economic Index by Anthropic, under CC BY. Wage and benefit data from the U.S. Bureau of Labor Statistics.
Now put a system beside the salary
The role cost calculator sets a loaded salary against the running cost of an AI system doing the same work, with the build counted in year one.
BLS wages May 2025 · O*NET 31.0 · Anthropic Economic Index